g

GE Digital Energy

UPS Topologies for

Large Critical Power Systems (> 500 KVA)
As presented at the 13th Annual Power Quality Exhibition & Conference, October 30, 2002

Dr. Edward R. Furlong

GM Technology
2018 Powers Ferry Road
Atlanta, GA 30339

GE Proprietary Information
October 2002

UPS TOPOLOGIES FOR LARGE CRITICAL POWER SYSTEMS (>500 KVA)

Dr. Edward R. Furlong, GM Technology
GE Digital Energy, Atlanta, GA
With the increasing demand of uninterrupted and high quality power for critical loads, high power
(>500KVA) UPS systems are becoming more and more popular. In order to properly design the critical
power system, the type of UPS and amount of redundancy must be matched to the nature of the load, the type
of power distribution, the quality of local power, and the required reliability. This requires a general
understanding of the performance of different UPS topologies and a guide to the trade-off analysis. In this
article, the four popular UPS topologies for high power (>500KVA) applications are described and a
comparative study of their response to grid-based PQ events is provided. A thorough analysis of the system
design, desired reliability, and various site constraints using the guidelines presented here will help ensure
that the customer receives the correct UPS topology.

namely normal, stored energy and bypass.

Under normal mode, the load is continuously
supplied by a rectifier-inverter combination, which
carries out the double conversion (AC-DC, then DCAC). The UPS goes into stored energy mode when
the AC input to the system fails or goes out of the
specified tolerance range. The inverter and the
battery continue to support the load under this mode.
The UPS runs in stored energy mode until the stored
energy is exhausted or until AC input returns to the
specified tolerance level.

Introduction
During the late 1990s, the rise of internet-based
companies those heavily dependent on servers and
other computer loads resulted in an explosion in
demand for electrical infrastructure capable of
protecting delicate equipment from outages (as short
as 1.5 cycles). While these particular companies
have suffered a recent downturn, the need for
conditioned power continues due to the increased
dependence on computer-controlled equipment in
fields such as medicine, bio-technology, and
semiconductor manufacturing. Even old economy
processes like coating, painting, and machining now
have a single computer controlling the finishing
operation on millions of dollars worth of finished
goods (such as the coatings on gas turbine blades).

This type of UPS is generally equipped with a static

bypass switch, allowing instantaneous transfer of the
load to the bypass AC input. This switch is used in
the event of a UPS internal malfunction, load current
transients (in-rush, or fault clearing), prolonged
overloads, or at the end of battery back-up (autonomy
time). However the presence of a bypass implies that
the input and output frequencies and phase must be
identical, and the voltage coordinated. The UPS is
synchronized with the source of AC bypass supply to
allow a transfer to bypass without any interruption.

Uninterruptible Power Supplies (UPSs) are used to

improve power source quality as well as protect these
critical loads against disturbances, such as frequency
shifts, voltage spikes and interruptions.1 There are
three major types of UPSs as specified by IEC
standard 62040-3.
Passive Standby (IEC 62040-3.2.20)
Line Interactive (IEC 62040-3.2.18)
Double Conversion (IEC 62040-3.2.16)
These terms define how the UPS delivers power to
the critical load. There are two popular styles of Line
Interactive UPS: Static and Rotary. The static line
interactive topology is popularly known as Delta
Conversion topology and the rotary line interactive
topology is termed as Rotary T topology in this
document.

Bypass AC
source

Normal Mode
Normal AC
source
Rectifier/
Charger

Double Conversion Topology

The most common type of large UPS today is the
Double Conversion UPS, shown in Figure 1. In this
topology, an inverter is connected in series between
the AC input and the load, with power for the load
flowing continuously through the inverter. Three
operating modes are possible in this topology,

Inverter
Critical
Load
Stored Energy Mode

Fig. 1. Block diagram of Double Conversion Topology

1
GE Proprietary Information

Bypass
Mode

Static Bypass
Switch

Rotary T UPS topology

The most common line-interactive topology is the
Rotary T, illustrated in Figure 3. Under normal
operating conditions (i.e. when the mains voltage is
within the accepted range), the synchronous motor
draws a compensating current so as to maintain a
constant load voltage level. Also during normal
operation, kinetic energy is stored in the flywheel (or
induction coupling). When the mains fail, the stored
flywheel energy is extracted to support the load.
Flywheel inertia (or induction coupling speed range)
is sized to support the full load operation for a
specific period (typically a few seconds), during
which time the backup diesel generators are started.

At the input of this UPS there is a rectifier, which

performs the AC-DC conversion. A 6-pulse rectifier
with a 5th harmonic filter reflects less than 7% input
current distortion. 12-pulse rectifiers or IGBTs
reflect even less, although at reduced levels of
efficiency, simplicity and reliability. While lower
distortion is desired, it is rarely needed and 7% THD
is sufficient to prevent any disruption to adjacent
loads or to the incoming utility feed for most
applications.
Delta Conversion Topology
The static line-interactive topology is termed Delta
Conversion, and shown in Figure 2. This is a line
interactive UPS topology with active seriesparallel
power conditioning capability. The sinusoidal output
voltage regulation capability results in low input
current and output voltage THD, in back up as well
as in standby mode. The series converter (mains side)
can be operated as a current source and the parallel
converter (Load side) as voltage source. In this case
the battery charging is done through the series
converter. Alternatively the parallel converter can be
operated as a current source and the series converter
as a voltage source under standby mode. In stored
energy mode, the parallel converter changes its
operating mode from current source to voltage
source. The static switch is switched off during stored
energy mode to prevent any power flow from the
battery to the mains. An advantage of this line
interactive topology over the double conversion
topology is its increased overall efficiency.

Special construction of the synchronous motor allows

this topology to provide good harmonic
compensation of the line current. Bi-directional
power converters are used to control power flow
between the dual winding synchronous
motor/generator and the flywheel synchronous
motor/generator.
The buffer inductor consists of two inductively
coupled windings. The impedance of the isolation
(mains side) is high (~49%) while the coupling
winding has a low impedance (~1%). The mutual
inductance between the two windings is equal to the
sub-transient reactance of the generator. The isolation
winding provides high impedance de-coupling
between the input and output of the UPS while the
generator regulates the voltage. The mutual
inductance between the windings offsets any voltage
variations at the output of the generator.
The main advantage of this topology is the ability to
clear large faults or start motors without transfer to
bypass. It is also favored for its compact size and
relatively high efficiency.

The input power factor and the line current

harmonics can be controlled in Delta Conversion.
The power factor is typically controlled to be close to
unity and the reflected line current distortion is
controlled to near 1% in many instances.
Bypass AC
source
Mains

Bypass Mode
Static Bypass
Switch
Series xf

Bypass switch

Normal Mode
Parallel
Converter

Static
switch

Series
Converter
Battery

Critical
Load

Mains

Input
disconnect

Stored
Energy
Mode

Normal mode

Buffer Inductor

Output
disconnect

Bi-directional
Power Converter

Synchronous
Motor/Generator

Fig. 3. Block diagram of Rotary T UPS Topology

Critical
Load

Stored Energy
mode

Induction
Coupling /
Flywheel

Fig. 2. Block diagram of Delta Conversion Topology

GE Proprietary Information

Bypass mode

Bypass AC
source

large loads to the grid system. These are also caused

by fault conditions at various points in the AC
distribution system. The typical duration of voltage
sags and swells at the grid system is 0.5 sec.

Passive Stand-By
In this topology, the system monitoring and control
unit continuously monitors the utility voltage and
frequency to determine if they are within the
specified limits. If the utility source parameters are
within +/-10%, the load is directly fed from the utility
mains through the fast acting static switch. During
this condition (standby mode) the power inverter
remains energized but does not supply any power to
the load. The battery charger maintains the voltage
level of the batteries during this time.

In Double Conversion and Rotary T topologies, the

output is very well isolated from the input side
voltage sags and swells. For the Delta Conversion
topology with a voltage swell >15%, the load voltage
experiences the swell until the static switch is
commutated off. Natural commutation at current zero
crossing can lead to maximum commutation time of
cycle. In the Passive Standby topology input sag /
swell directly gets passed to the load for short
durations (~15ms).

In the event of any power failure at the mains or if

the mains power parameters go out of the specified
limit, the system control turns off the fast acting
static switch and turns on the inverter. Under this
condition (Stored energy mode) the inverter supplies
the load power by converting energy from the
battery.

Line frequency variation: For satisfactory operation

of most of the computer equipment, frequency
variation should be limited to +/-0.5Hz. UPS
topologies other than Double Conversion cover
mains frequency variations by tapping into the stored
energy resource (Flywheel or Battery). Double
Conversion topology can maintain the load frequency
variation within +/-0.1%, without tapping the stored
energy. In the line-interactive topologies as well as
in the Passive Standby topology, the load experiences
equal frequency variation as the mains. When this
variation reaches unacceptable levels, the control
circuit isolates the load from the mains through the
input disconnect. This results in a discharge of stored
energy until the utility source is back within
tolerance. If this type of frequency variation is very
common, the battery/ flywheel discharges frequently,
resulting in possible early battery failure or a high
level of nuisance starts on the genset. This may also
result in a decreased amount of stored energy during
an actual power interruption.

Operation of the UPS under stored energy mode is

the same as that of a double conversion topology.
The major drawback of this topology is that it does
not condition the utility power in any way during
standby mode operation. Thus, in this topology the
UPS transfers to stored energy mode very often.
However, a major advantage of this topology is its
simplicity of control, high efficiency and low cost.
Bypass AC
source

AC
Mains

Bypass Switch

static
switch

Bypass mode

Normal mode

Critical
Load
Battery
Charger

System
monitoring
and control

Battery

Input voltage unbalance and single phasing. The

ANSI (American National Standards Institute)
recommended limit of voltage unbalance in power
distribution system is 3%. Approximately 98% of US
distribution systems have unbalances less than or
equal to 3%. Single phasing can be considered as a
special case of voltage unbalance equal to 100%.

Stored Energy mode

AC
Capacitor

Inverter

Fig. 4. Block diagram of Passive Standby UPS Topology

Response to Power Quality Events

The four UPS topologies described in the previous
section were simulated2 in Saber, a commercially
available computer modeling tool. The models did
not include all components. However, they were
suitable to simulate the macro behavior of the UPS in
response to typical grid-based power quality events
with sufficient detail to make comparative
performance assessments.
Line voltage sag and swell. Line voltage sags and
swells happen during the application or removal of

All the UPS topologies except passive standby

protect the load from typical input voltage unbalance
(<3%). For higher levels of unbalance, the losses in
the Rotary T topology increase.
When there is single phasing at the incoming utility
feed, all of the UPS topologies draw from the stored
energy source. Double Conversion and Delta
Conversion topologies transfer the load to stored
energy instantaneously without any disturbance. In

3
GE Proprietary Information

and swells at the input. Typical time of operation of

this switch is 200ms. Slow operation of the breaker
does not affect the performance of the UPS, as the
UPS can isolate load voltage from mains for short
duration (about 1-sec). The breaker operation time
varies depending upon the rms ac as well as the DC
content of the current through it. Passive standby
UPS has one forced commutated static switch at the
input. Typical commutation time of this switch is
4ms ( cycle).
The total input disconnection time is the sum of
switch commutation time and fault detection time.
The fault detection time depends on the control
circuit, and is typically on the order of cycle.

the Rotary T topology, the load voltage experiences a

~20% dip for ~200ms. In Passive Standby topology,
the UPS load experiences the single phasing for a
minimum of 4ms.
Voltage waveform distortion at input. In the
double conversion topology the load side voltage is
completely independent of the mains waveform
distortion. In delta conversion topology the parallel
converter is controlled as a voltage source and
regulates the load voltage to achieve a good voltage
THD on the order of 3%, irrespective of the input
voltage THD. In case of Rotary T topology, the
isolation inductor acts as a filter to the input voltage
distortion. However, the isolation inductor does not
filter any sub-harmonics in input voltage and passes
on the distortion to the load side. A passive standby
UPS does not protect the load from utility voltage
distortion.

ITI (CBEMA) compliance of UPS topologies:

The Information Technology Industry Council (ITI)
Curve3, formerly the Computer & Business
Equipment Manufacturers Association (CBEMA)
Curve, describes an AC input voltage envelope,
which typically can be tolerated by most Information
Technology Equipment. The curve describes both
steady state as well as transient conditions. It may be
noted that this curve does not address frequency
variation and waveform distortion directly, except for
some guidelines as to energy absorption (80 J).
Figure 5 shows the voltage envelope offered to the
load by double conversion and Rotary T topologies.
Overlaid on this envelop is the CBEMA or ITIC
boundary, which is the general performance standard
required by computer power supplies and is roughly
similar to the SEMI F47 curve used in the
semiconductor industry.

Input switch commutation time. Among the four

topologies discussed here, all but double conversion
have a switch at the input which dictates the
performance of the UPS.
Delta conversion topology has a line commutated
thyristorized switch at the input, which is
commutated naturally at the current zero crossing.
Thus the time delay for the switch to isolate the UPS
from input side fault depends on the type and instant
of occurrence of the fault. This static switch operates
within 1ms for faults like input short circuit or input
open circuit provided input power factor is
maintained at unity. For other faults, the maximum
time is cycle. Rotary T topology has one motorized
breaker at the input. Operation of this breaker is
relatively infrequent as the UPS isolates the output
from input side faults very well, including large sags

Table 1: Overview of system responses

Double Conversion
Voltage
Sag

Voltage
Swell

Delta Conversion

Load voltage is regulated within

Load voltage is regulated
+/-0.5% at Steady State, and +/- within +/-0.5% at Steady State,
5% in dynamic state (1 power and +/-5% in dynamic state (1
cycle)
power cycle)
Load voltage is regulated within
Voltage surge (>140%) is
+/-0.5% at Steady State, and +/experienced at the load for
5% in dynamic state (3 power
short duration (~1 cycle)
cycles)

Frequency Frequency variation is restricted

variation
to +/-0.5% at load

No control on line frequency

variation

Voltage
waveform
distortion

Load voltage waveform is not

distorted by line voltage
distortion

Line voltage distortion is

corrected by the parallel
converter at the load point

Voltage
unbalance

Load does not see the input

voltage unbalance

Load does not see the input

voltage unbalance

4
GE Proprietary Information

Rotary T

Passive Standby

Load voltage is regulated

During normal operation up
within +/-1% at Steady State,
to 10% voltage sag is
and +/-5% in dynamic state
experienced by the load
(200ms)
Load voltage is regulated
Voltage surge is experienced
within +/-1% at Steady State,
at the load for short duration
and +/-5% in dynamic state
(~1cycle)
(200ms)
No control on line frequency No control on line frequency
variation
variation
Line voltage distortion is
corrected to a large extent
by the synchronous motor
(short circuit ring) at the load
point
Line voltage unbalance is
corrected to a large extent
by the synchronous motor
(low negative sequence
impedance) at the load point

Line voltage distortion is

passed on to the load

Line voltage unbalance is

passed on to the load

boundaries are dependent on the input static switch

commutation time. For delta conversion and passive
standby topology the maximum commutation time
for the static switch is assumed to be 0.5 cycles and
4ms respectively. From this plot, it is seen that many
operating points in these topologies can go out of the
CBEMA boundary. This is especially true for input
voltage swells.

150

Rotary
Double Conversion

100

CBEMA Boundary

50

# of Power Cycles
0
0.001

0.1

10

1000

100000

Fig. 5. Operating boundaries of Double Conversion and

Rotary-T

Figure 7 shows the sensitivities of a Double

conversion UPS and Rotary T UPS to typical voltage
sag/swell events in a power distribution system.
Double conversion UPS regulates the load voltage to
+/-5%, with maximum dynamic time of 3 cycles.
Output is maintained within +/-2%, instantaneously
for grid voltage sag and swell of 15% or less. (In the
above plot, instantaneous operation is considered as
1ms, for presentation only). For higher sag and swell
also the output voltage is regulated to +/-2% at steady
state. However, during dynamic condition the output
voltage can deviate by maximum of +/-5%. The time
period for this deviation may vary from 1 cycle to 3
cycles.

% of Nominal
Voltage

200

150

Passive Standby
Delta Conversion

100

CBEMA Boundary

50

# of Power Cycles
0
0.001

0.1

10

1000

100000

Fig. 6. Operating boundary of Delta conversion and Passive

Standby

In the double conversion topology, output voltage

reaches steady state within a maximum of three
power cycles. At steady state, the output voltage is
tightly regulated within +/-2% of the nominal
voltage. During dynamic conditions (duration less
than three power cycles), the voltage does not go
beyond +/- 10% of the nominal.

The rotary-T topology UPS regulates the load voltage

to +/-5% instantaneously for grid voltage variation
within +20% to -50%. For higher sag and swell the
output voltage is also regulated to +/-5% at steady
state. However, during a dynamic condition but
before the input motorized breaker is opened, the
output voltage can deviate by maximum of +/-10%.
The time period for this deviation may vary from 10
cycles to 20 cycles.
250

200

The rotary T topology operating boundary looks very

similar to that of the double conversion topology,
except that, its dynamic response time is slower.
Thus larger voltage variation at the output can be
observed for longer duration. However, in this
topology, all operating points, including dynamic
state, fall within the CBEMA Curve.

150

100

50

Figure 6 shows the voltage envelope offered to the

load by delta conversion and passive standby
topologies. In these topologies, the operating

Rotary T
PQ events

Double
Conversion
# of Power Cycles

0
0.001

0.1

10

1000

Fig. 7. PQ events with Double conversion and Rotary T

5
GE Proprietary Information

Figures 7 and 8 are scatter plots which show the

results of simulations aimed at illustrating how
various UPS topologies are able to protect a typical
semiconductor fabrication laboratory from grid
voltage sags and swells. All round points on these
plots indicate the practical data collected from the
site (Source: EPRI Distribution Power Quality study).

% of Nominal Voltage

% of Nominal
Voltage

200

100000

Tradeoff Analysis
In order to properly design the critical power system,
the type of UPS and amount of redundancy must be
matched to the nature of the load, the type of power
distribution, the quality of local power, and the
required reliability. These attributes vary by location
and customer, so it is important not to prejudge the
solution going strictly by past experience. The
objective of this section is to present some of the
tradeoffs that can be used to constrain the myriad of
options available to the design engineer or customer.

250

150

100

% of Nominal Voltage

Passive Standby
200

Delta
Conversion

PQ events

50

The first step is to identify the size of the UPS by

choosing the type of distribution system. There are
three general approaches:
1. A distributed architecture where the UPS is
placed directly next to each critical load.
2. A centralized architecture where all UPS are
grouped together
3. A grouped approach where a handful of UPS
systems are located strategically around the
facility.

# of Power Cycles
0
0.001

0.01

0.1

10

100

1000

10000 100000

Fig. 8. PQ events with Delta conversion and Passive Standby

Figure 8 shows the sensitivities of Delta conversion

UPS and Passive Standby UPS to typical voltage
sag/swell events in a power distribution system. A
delta conversion ups regulates the load voltage to +/2% instantaneously for mains voltage of +/-15%. For
higher sags and swells the output voltage is regulated
to +/-2% at steady state. However, during dynamic
condition, before the input static switch is
commutated off, the output voltage experiences the
input voltage swell in a reduced scale. The time
period for the static switch commutation (including
fault detection time) may vary from 0.3 cycles to 0.8
cycles, depending upon the fault condition.

For large UPS systems, it is generally accepted that

the distributed architecture leads to a system that is
difficult and costly to maintain. The decision of
whether to group or centralize is often more subtle,
and is partly dependent on the type of UPS and the
layout of the site. However, in either case, the UPS
system is most reliable when kept in blocks of 2-5
units, provided the units are connected in an
architecture like Redundant Parallel ArchitectureTM
(RPATM) as shown in Figure 9. This guideline is the
result of a classic tradeoff between the greater
dynamic response and proven fault tolerance of
multi-unit systems with the higher theoretical MTBF
of single-unit systems.

The passive standby UPS passes the grid voltage to

the load without conditioning for sags/swells less
than 10%. For higher sags/swells, the input static
switch is commutated off and the output voltage is
regulated to +/-2% at steady state. However, until the
static switch is commutated off, the output voltage
experiences the input voltage sags and swells
unattenuated. The time period for the static switch
commutation may vary from 0.2 cycles to 0.3 cycles.

With the understanding that 2-5 units comprise the

design target, simply knowing the power
requirements of the facility will constrain the unit
size. Consider a 5-MW facility, common in the
semiconductor industry. The UPS size would need to
be 1-2 MW. On the other hand, a 1-MW data center
could use a group of 200-500 kVA units.

Figures 7 and 8 indicate that whereas Double

Conversion and Rotary T topologies protect the load
satisfactorily, the load voltage may experience a
voltage swell exceeding the CBEMA limit if it is
protected by Delta conversion or Passive Standby
topology UPS. The probability of such an event is
0.8%, as calculated from the above sample data.
Improvements to this failure rate require
sophisticated control algorithms for the switching,
similar to those found in PDUs. However, this can
lead to different failure modes and care is advised.

Fig. 9 Redundant Parallel Architecture (RPA) implementation.

6
GE Proprietary Information

500KVA

1MVA

1.3MVA

Phase reversal

2MVA

Frequency shift

Input short
circuit
Voltage
unbalance

Mid-Range

Small

Cost-Performance
Trade-off

Rotary
Line-Int

Interruption

Large
500KVA

Output short
Circuit

Cost drives rotary

Static UPS

Double
Conversion

Frequency
Transients

Rotary UPS

Cost drives static

Mega
1MVA

1.3MVA

2MVA

Waveform
Distortion

Fig. 10. Tradeoff between Static and Rotary systems.

0.01

For the large semiconductor facility, it may make

sense to use a rotary UPS because of the larger unit
size, higher efficiency, and lower installed cost (on a
$/kW basis). However, the smaller data center is
more likely to choose a static UPS because of the
better dynamic response, lower maintenance, and
general ease of use of a static UPS system. This
tradeoff is illustrated in Figure 10.
The last item to consider is the local power quality.
While the response to general faults is indicated in
Table 1, it is sometimes more instructive to consider
the covered faults along with their estimated
likelihood at the proposed facility, as shown in Figure
11. Some events, like waveform distortion, can occur
monthly. Others, like a phase reversal, may occur
only once in a 1000 yrs.

0.1

10

100

1000

10000

Mean Time Between Event (yrs)

Fig. 11 Sample probability of events and ability of different

topologies to clear.

Double
Conversion

Rotary
Line-Int.

Passive
Off Line
Standby

N+N

N+1

Installed
Cost

N+0

Process

Batch
Process

Data
Centers

Assembly

While that is indeed very unlikely, consider that this

implies a 0.1% chance of failure each year,
essentially bounding the reliability that can be
achieved at the site. This further serves to narrow
down the choice of UPS by bracketing the desired
level of reliability as shown in Figure 12.

99%

99.9%

99.99%

99.999% 99.9999% 99.99999%

Reliability
Fig. 12 Asymptotic cost analysis of different topologies as a
function of reliability using probabilities from Fig. 11. While no
amount of redundancy can produce a 6-9s system with passive
standby, not all customers require this or even N+N systems.4

The analysis used to derive Fig. 12 is asymptotic in

nature, meant to capture the basic influence of the
event probability in Fig. 11. The techniques are
simplified from Ref. 5 & 6 and assume that any PQ
event that slips past the chosen topology will cause a
24-hr outage.

events will sneak past the system, even one that is

switch-based.
Summary
There are several important UPS topologies available
today for large critical power systems. Because of the
degree of isolation in a Double-Conversion UPS, this
topology offers superior protection from line-side PQ
events. The Rotary-T topology provides a balance
between line-side and load-side protection, although
with reduced levels of isolation and dynamic
response as compared to Double Conversion. The
Delta Conversion topology has some advantages with
regard to improved efficiency over the Double
Conversion. However, it has the potential to allow
voltage surges and some other PQ events to pass

Notice that for the assumed event frequency, no

amount or configuration of passive standby UPSs
can reach the level of reliability afforded by a
properly designed and installed double conversion
UPS system. However, there are many applications
where the lower cost is more important to the
customer, and they will simply accept the lower
reliability as a business decision. In fact in most
cases with N+0 UPS designs, it is more likely that the
UPS itself will fail than one of the power quality

7
GE Proprietary Information

Passive
Standby

Output single
phasing

through to the load. In addition, the manner in which

the UPS mitigates PQ events often results in battery
drainage, which in turn leads to reduced battery life.
This last issue also applies to Passive Standby, which
does little to no power conditioning, and must switch
to battery power whenever the input voltage is out of
tolerance.

References
1. Prasad P, et al., Topologies of Uninterruptible
Power Supplies used in Critical Power Systems,
GE Digital Energy Report No. 11, 2001.
2. Prasad P, et al , Sensitivities of various UPS
systems to grid-based power quality events, GE
Digital Energy Report No. 12, 2001.

Static Double Conversion topology and Static Line

interactive topology (Delta Conversion) are viable
solutions in lower power ranges (unit size <1MVA).
With large unit power ratings and availability at
medium voltage, the Rotary-T topology is the
preferred solution for the larger installations. The
passive standby solutions are viable for customers
looking to avoid the business interruption caused by
only the most common PQ events.

3. The Information Technology Industry Council

(ITI) Curve, formerly the Computer & Business
Equipment Manufacturers Association (CBEMA),
is located at www.itic.org/technical/iticurv.pdf
along with a detailed application note.
4. IEEE Gold Book (1997) - IEEE Std 493-1997,
IEEE Recommended Practice for the design of
Reliable Industrial and Commercial Power
Systems

A thorough analysis of the system design, desired

reliability, and various site constraints using the
guidelines presented here will help ensure that the
customer receives the correct UPS topology.

5. Wang et al, Apply simulation technique via

reliability block diagram to gold book standard
network, submitted to IEE Gold Book Working
Group

Acknowledgements
The numerical analysis of the topology response was
led by Bansi Phansalker and his team at the Jack F.
Welch Technology Center, part of GE Global
Research-Bangalore. The interpretation of results
was done in conjunction with Robert Delmerico at
GE Corporate Research-Niskayuna.

6. Wendai Wang and James Loman On Reliability

Modeling and Analysis of Highly-reliable, Large
Systems, Proceedings of the 48th Annual
Reliability and Maintainability Symposium, 2002,
pp. 456-459.

8
GE Proprietary Information